AP® Psychology · Science Practices: Research Methods and Data
2.C Why Correlation Is Not Causation
A correlation tells you two variables rise and fall together, but it cannot tell you why. The link might be a third variable causing both, or the direction of cause and effect might be reversed.
A correlation is just a pattern: two variables move together, either in the same direction or in opposite directions. The number describing it, the correlation coefficient, says nothing about mechanism. It cannot distinguish between three very different stories: variable A causes variable B, variable B causes variable A, or some third variable C is causing both A and B independently. All three stories can produce the exact same scatterplot, same strength, same direction.
The third-variable problem is the most common trap. Ice cream sales and drowning deaths both climb in summer, not because one causes the other, but because warm weather drives both: it pushes people toward ice cream and toward swimming, and more swimming means more drowning. Once you measure temperature and control for it, the direct link between ice cream and drowning disappears. The directionality problem is different: even when only two variables are involved, a correlation cannot tell you which one is the cause. Video game hours and low grades might mean gaming drags grades down, or struggling students may retreat into gaming, the arrow could point either way and the data looks identical either way.
This is why the science practices unit treats correlational findings as a starting point, not an answer. Spotting a correlation is step one of research; the next step is designing a study, ideally an experiment with manipulation and random assignment, that can isolate cause. Once a suspected third variable is actually measured and statistically controlled, a researcher can check whether the original correlation shrinks or vanishes, which is exactly how warm weather gets identified as the true cause behind ice cream and drowning.
Key terms
- Correlation
- A measured relationship in which two variables tend to rise or fall together, described by strength and direction.
- Third-variable problem
- A hidden factor independently causes both measured variables, creating a correlation between them with no direct link.
- Directionality problem
- A correlation between two variables cannot reveal which one is the cause and which is the effect.
- Confound
- Any outside variable that offers an alternative explanation for an observed correlation.
Watch it explained
- How it worksHow does a third variable produce a correlation between two things it separately causes?
- How it worksWalk me through the directionality problem with the arrow pointing both ways.
- Worked exampleShow me three causal diagrams that all produce the same scatterplot.
- Going furtherWhat changes about a correlational finding once the third variable is measured?
Check yourself
End of topic.